OpenAI’s Q* – What Do We Know?
Did OpenAI fire Sam Altman due to a mysterious AI model? Word has it that OpenAI made a breakthrough earlier this year, resulting in Q* (pronounced Q star”), an AI able to solve grade-school math problems.
Why is middle-school math such a big deal that it caused several staff members to write a letter to OpenAI’s board warning them the breakthrough could threaten humanity? Could the warning be one of the reasons the board chose to fire Sam Altman, who is now back at OpenAI as CEO?
The ability to solve basic math is a giant leap forward from existing AI models that need help with logical reasoning about abstract concepts outside the data they trained on. Language models struggle as they lack logical reason.
Q*’s name implied it could combine the deep learning techniques powering ChatGPT with human-programmed rules, symbolically connecting the two sides of the brain, knowing some things from experience while still being able to reason about facts.
The day before Altman was fired, he discussed an advancement that would “push the veil of ignorance back and the frontier of discovery forward.” Some OpenAI staff believed Q* could be a breakthrough in the company’s goal of achieving artificial general intelligence or autonomous systems that surpass humans in most economically viable tasks.
Following Altman’s termination, over 700 of OpenAI’s 770 staff signed a letter on Monday threatening to leave for Microsoft if he wasn’t reinstated. He returned Tuesday night.
“What Do We Know?” Apparently very little, at least not enough.
I understand the advancement of achieving rule-based inference, but that’s not totally new, and doesn’t come close to an threat to humanity’s existence!